%matplotlib widget
import pandas as pd
import numpy as np
from sklearn import datasets
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA
from FisherDA import MultipleFisherDiscriminantAnalysis as MDA
from sklearn.decomposition import PCA
from sklearn.svm import LinearSVC
from sklearn.svm import SVC
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
import matplotlib
import matplotlib.pyplot as plt
from matplotlib.colors import ListedColormap
from mpl_toolkits.mplot3d import Axes3D
from IPython.display import display
from sklearn.preprocessing import LabelEncoder
from sklearn.metrics import precision_score, recall_score, fbeta_score

#importiamo il dataset
adult = pd.read_csv('adult.data', header=None,na_values=" ?")
adult_columns = ['age', 'workclass', 'fnlwgt', 'education', 'education-num', 'marital-status', 'occupation', 'relationship','race','sex','capital-gain','capital-loss','hours-per-week','native-country','income']
adult.rename(columns={k:adult_columns[k] for k in range(len(adult_columns))}, inplace=True)

display(adult)

adult.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 32561 entries, 0 to 32560
Data columns (total 15 columns):
 #   Column          Non-Null Count  Dtype 
---  ------          --------------  ----- 
 0   age             32561 non-null  int64 
 1   workclass       30725 non-null  object
 2   fnlwgt          32561 non-null  int64 
 3   education       32561 non-null  object
 4   education-num   32561 non-null  int64 
 5   marital-status  32561 non-null  object
 6   occupation      30718 non-null  object
 7   relationship    32561 non-null  object
 8   race            32561 non-null  object
 9   sex             32561 non-null  object
 10  capital-gain    32561 non-null  int64 
 11  capital-loss    32561 non-null  int64 
 12  hours-per-week  32561 non-null  int64 
 13  native-country  31978 non-null  object
 14  income          32561 non-null  object
dtypes: int64(6), object(9)
memory usage: 3.7+ MB

#elimino tutte le righe con valori nulli e salvo le dimensioni
adult_b = adult.dropna()
M,N = adult_b.shape
adult['income'].value_counts()

income
<=50K    24720
>50K      7841
Name: count, dtype: int64

#mostro che la maggior parte dei dati eliminati guadagna meno di 50k(però non è un problema perchè la differenza è ancora sostanziale)
adult_b['income'].value_counts()

income
<=50K    22654
>50K      7508
Name: count, dtype: int64

#una distribuzione dell'età
plt.figure(figsize=(8, 4))
plt.hist(adult_b['age'], bins=25, edgecolor='black')
plt.title('Distribuzione dell\'età')
plt.xlabel('Età')
plt.ylabel('Frequenza')
plt.xticks(ticks=np.arange(16,92,5), labels=[f'{i}' for i in range(16,92,5)])
plt.yticks(ticks=np.arange(0,2750,250), labels=[f'{i}' for i in range(0,2750,250)])
plt.grid(False)
plt.show()

average_age = np.round((adult_b['age']*adult_b['fnlwgt']).sum()/adult_b['fnlwgt'].sum())
display("L'età media ponderata è: "+str(average_age)+" anni")

"L'età media ponderata è: 38.0 anni"

#trasformo le colonne object in category
adult_b=adult_b.astype({"workclass":'category', "education":'category', "marital-status":'category', "occupation":'category', "relationship":'category',"race":'category',"sex":'category',"native-country":'category',"income":'category'})      

#Visto che le classi income sono due (>50k, <=50k), per esse possiamo anche utilizzare un label encoding. Per le features adotteremo invece sempre il one-hot encoding
adult_work=adult_b.copy()
le=LabelEncoder()
le.fit(adult_work['income'].values)
adult_work['income']=le.transform(adult_work['income'].values)

#one-hot encoding
onehot_cols=['workclass', 'education', 'marital-status', 'occupation', 'relationship', 'race','sex','native-country']
adult_work_old=adult_work.copy()
adult_work=pd.get_dummies(adult_work, columns=onehot_cols)
adult_work=adult_work.astype(int)
col_to_move = adult_work.columns[6]
new_pos = 104
cols = adult_work.columns.tolist()
cols.insert(new_pos, cols.pop(cols.index(col_to_move)))
adult_work = adult_work[cols]

display(adult_work)

#mostriamo la relazione tra età e ore settimanali lavorative
cmap=ListedColormap(['blue', 'orange'])
plt.figure(figsize=(8, 4))
plt.scatter(adult_work_old['age'], adult_work_old['hours-per-week'], c=adult_work_old['income'],cmap=cmap, alpha=0.5)
plt.title('Relazione tra Età e ore lavorative settimanali')
plt.xlabel('Età')
plt.ylabel('ore lavorative settimanali')
handles = [plt.Line2D([0], [0], marker='o', color='w', markerfacecolor=cmap(0), markersize=10, label='<=50K'),
           plt.Line2D([0], [0], marker='o', color='w', markerfacecolor=cmap(1), markersize=10, label='>50K')]
plt.legend(handles=handles, title='income')
plt.xticks(ticks=np.arange(16,92,5), labels=[f'{i}' for i in range(16,92,5)])
plt.yticks(ticks=np.arange(0,110,10), labels=[f'{i}' for i in range(0,110,10)])
plt.grid(True)
plt.show()

#mostriamo la relazione tra guadagni da investimenti ed età
plt.figure(figsize=(8, 4))
plt.scatter(adult_work_old['age'], adult_work_old['capital-gain'], c=adult_work_old['income'], cmap=cmap, alpha=0.50)
plt.title('Età vs Guadagno di Capitale')
plt.xlabel('Età')
plt.ylabel('Guadagno di Capitale')
plt.legend(handles=handles, title='Income')
plt.xticks(ticks=np.arange(16,92,5), labels=[f'{i}' for i in range(16,92,5)])
plt.yticks(ticks=np.arange(0,110000,10000), labels=[f'{i}' for i in range(0,110000,10000)])
plt.grid(True)
plt.ion()
plt.show()

#mostriamo la relazione tra ore lavorative settimanali e il peso che hanno sui dati
plt.figure(figsize=(8, 4))
plt.scatter(adult_work_old['age'], adult_work_old['capital-loss'],  c=adult_work_old['income'], cmap=cmap, alpha=0.5)
plt.title('Età vs Perdita di Capitale')
plt.xlabel('Età')
plt.ylabel('Perdita di Capitale')
plt.legend(handles=handles, title='Income')
plt.xticks(ticks=np.arange(16,92,5), labels=[f'{i}' for i in range(16,92,5)])
plt.yticks(ticks=np.arange(0,5000,500), labels=[f'{i}' for i in range(0,5000,500)])
plt.grid(True)
plt.show()

#mostriamo la correlazione tra le variabili
adult_work.iloc[:, :-1].corr()

#separiamo le classi dalle features
x=adult_work.iloc[:, :-1].values
y=adult_work['income'].values

#generiamo il training
random_seed = 42  
test_p = 0.50  
X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=test_p, random_state=random_seed, shuffle=True)

#standardizzo
scaler_X = StandardScaler()
scaler_X.fit(X_train)
X_scaled = scaler_X.transform(X_train)

#applichiamo la PCA sia st che nonst
pca=PCA()
pca_nonst=PCA()
pca.fit(X_scaled)
pca_nonst.fit(X_train)    

#mostriamo la varianza spiegata
plt.figure()
plt.plot(np.insert(np.cumsum(pca.explained_variance_ratio_), 0, 0))
plt.title('x_scaled')
plt.xticks(ticks=np.arange(1,pca.n_components_+3,8), 
           labels=[f'PC{i}' for i in range(1, pca.n_components_ + 3,8)], rotation=45)
plt.xlabel('Principal components')
plt.ylabel('Cumulative explained variance (%)')
plt.grid()
plt.show()

plt.figure()
plt.bar(np.arange(pca_nonst.n_components_), pca_nonst.explained_variance_ratio_, bottom=np.insert(np.cumsum(pca_nonst.explained_variance_ratio_), 0, 0)[:-1])
plt.plot(np.cumsum(pca_nonst.explained_variance_ratio_), 'r')
plt.title('x_nonst')
plt.xticks(ticks=np.arange(1,pca_nonst.n_components_,10), 
           labels=[f'PC{i}' for i in range(1, pca_nonst.n_components_ + 1,10)])
plt.xlabel('Principal components')
plt.ylabel('Cumulative explained variance (%)')
plt.grid()
plt.show()

#scegliamo il numero di componenti principali in base alla varianza spiegata
explvar_p = 0.90
pca=PCA(explvar_p)
pca.fit(X_scaled)
pca_adult=pca.transform(X_scaled)

#trasformo i miei dati originali in quelli trasformati dalla PCA
features_scalate=scaler_X.transform(x)
Zpca = pca.transform(features_scalate)

df_pca = pd.DataFrame({'val': [pca.n_components_, np.round(pca.explained_variance_ratio_.sum()*100,decimals=2)]}, index=['n. PC', 'expl. Var. (%)'])
display(df_pca)

#mostriamo il grafico a dispersione
plt.figure(figsize=(12, 12))
plt.scatter(pca_adult[:, 0], pca_adult[:, 1], c=y_train, cmap=cmap, alpha=0.5)
plt.title('ADULT - SCORE GRAPH - 2D')
plt.xlabel('PC1')
plt.ylabel('PC2')
plt.legend(handles=handles, title='income')
plt.grid()
plt.show()

fig = plt.figure(figsize=(12, 12))
ax = fig.add_subplot(111, projection='3d')
ax.scatter(pca_adult[:, 0], pca_adult[:, 1],pca_adult[:,2], c=y_train, cmap=cmap, alpha=0.5)
plt.title('ADULT - SCORE GRAPH - 3D')
ax.set_xlabel('PC1')
ax.set_ylabel('PC2')
ax.set_zlabel('PC3')
plt.grid()
plt.show()

#mostriamo la correlazione tra le componenti principali e le features
plt.figure(figsize=(18, 15))
for i in range(pca.n_components_):
    plt.plot([0, pca.components_[0, i]], [0, pca.components_[1, i]], label=adult_work.iloc[:, :-1].columns[i])
    plt.scatter(pca.components_[0, i], pca.components_[1,i], c='k')
plt.legend(bbox_to_anchor=(1.05, 1), fontsize='x-small')
plt.title('ADULT - LOADING GRAPH 2D')
plt.xlabel('PC1')
plt.ylabel('PC2')
plt.grid()
plt.show()

fig = plt.figure(figsize=(18, 15))   
ax = fig.add_subplot(111, projection='3d')
for i in range(pca.n_components_):
    ax.plot([0, pca.components_[0, i]], [0, pca.components_[1, i]], [0, pca.components_[2, i]])
    ax.scatter(pca.components_[0, i], pca.components_[1, i], pca.components_[2, i], c='k')
plt.title('ADULT - LOADING GRAPH 3D')
ax.set_xlabel('PC1')
ax.set_ylabel('PC2')
ax.set_zlabel('PC3')
plt.grid()
plt.show()

#tramite dei barplot mostriamo la correlazione tra le componenti principali e le features
plt.figure(figsize=(18, 15))
plt.bar(np.arange(0,pca.n_components_,1), pca.components_[0, :-25])
col_n=[None]*pca.n_components_
for i in range(pca.n_components_):
    col_n[i]=adult_work.iloc[:, :-1].columns[i]
plt.xticks(ticks=np.arange(pca.n_components_), labels=col_n,rotation=90, fontsize='x-small')
plt.title('PC1')
plt.grid()
plt.show()

plt.figure(figsize=(18, 15))
plt.bar(np.arange(0,pca.n_components_,1), pca.components_[1, :-25])
col_n=[None]*pca.n_components_
for i in range(pca.n_components_):
    col_n[i]=adult_work.iloc[:, :-1].columns[i]
plt.xticks(ticks=np.arange(pca.n_components_), labels=col_n,rotation=90, fontsize='x-small')
plt.title('PC2')
plt.grid()
plt.show()

#mostriamo la separazione tra classi dell'MDA
mda=MDA()
mda.fit(X_train, y_train)
Zmda = mda.transform(x)

zero=np.zeros(Zmda.shape[0])
fig1, axs1 = plt.subplots(1, 2, figsize=(14, 10))
axs1[0].scatter(Zmda[:, 0],zero, c=y, cmap=cmap, alpha=0.50)
handles = [plt.Line2D([0], [0], marker='o', color='w', markerfacecolor=cmap(0), markersize=10, label='<=50K'),
           plt.Line2D([0], [0], marker='o', color='w', markerfacecolor=cmap(1), markersize=10, label='>50K')]
axs1[0].legend(handles=handles, title='income')
axs1[0].set_title('MDA')
axs1[0].grid()
axs1[0].set_xticks(np.arange(Zmda[:, 0].min()-1, Zmda[:, 0].max()+1))  
axs1[1].scatter(Zpca[:, 0], Zpca[:, 1], c=y, cmap=cmap, alpha=0.50)
axs1[1].legend(handles=handles, title='income')
axs1[1].set_title('PCA')
axs1[1].grid()

#applichiamo la MDA al campione della PCA
mda_pca=MDA()
mda_pca.fit(Zpca, y)
Zmda_pca = mda_pca.transform(Zpca)

fig1, axs1 = plt.subplots(1, 2, figsize=(14, 5))
axs1[0].scatter(Zmda[:,0].flatten(), y, c=y, cmap=cmap, alpha=0.15)
axs1[0].set_title('MDA (senza PCA)')
axs1[0].grid()
axs1[0].set_xticks(np.arange(Zmda[:, 0].min(), Zmda[:, 0].max()))
axs1[1].scatter(Zmda_pca.flatten(), y, c=y, cmap=cmap, alpha=0.15)
axs1[1].set_title('MDA (con PCA)')
axs1[1].grid()

#calcolo la precisione con cui il metodo LDA predice dei dati(no PCA)
lda =LDA()
lda.fit(X_train, y_train)

y_pred=lda.predict(X_test)
y_pred_proba=lda.predict_proba(X_test)
y_pred_df= pd.DataFrame({'Pred. Class': y_pred, 
                          'P(Class 0) - %': np.round(y_pred_proba[:, 0] * 100, decimals=2), 
                          'P(Class 1) - %': np.round(y_pred_proba[:, 1] * 100, decimals=2)})
scores_dict = {'Training Set': np.round(lda.score(X_train, y_train)*100, decimals=2), 'Test Set': np.round(lda.score(X_test, y_test) * 100, decimals=2)}
scores = pd.DataFrame(scores_dict, index=['Accuracy'])

display(scores)
display(y_pred_df)

#calcolo la precisione del metodo LDA con PCA
lda_pca =LDA()
lda_pca.fit(pca_adult, y_train)

y_pred=lda_pca.predict(pca.transform(scaler_X.transform(X_test)))
y_pred_proba=lda_pca.predict_proba(pca.transform(scaler_X.transform(X_test)))
y_pred_df= pd.DataFrame({'Pred. Class': y_pred, 
                          'P(Class 0) - %': np.round(y_pred_proba[:, 0] * 100, decimals=2), 
                          'P(Class 1) - %': np.round(y_pred_proba[:, 1] * 100, decimals=2)})
scores_dict = {'Training Set': np.round(lda_pca.score(pca_adult, y_train)*100, decimals=2), 'Test Set': np.round(lda_pca.score(pca.transform(scaler_X.transform(X_test)), y_test) * 100, decimals=2)}
scores = pd.DataFrame(scores_dict, index=['Accuracy'])

display(scores)
display(y_pred_df)

#calcolo la precisione del metodo SVM hard
C_hard = 1e10
loss = 'squared_hinge'
dual = False

lsvm_hard = LinearSVC(C=C_hard, loss=loss, dual=dual, random_state=random_seed)
lsvm_hard_pca = LinearSVC(C=C_hard, loss=loss, dual=dual, random_state=random_seed)

lsvm_hard.fit(X_train, y_train)
lsvm_hard_pca.fit(pca_adult, y_train)

df_lsvm_hard = pd.DataFrame({'acc. (no PCA)': [np.round(lsvm_hard.score(X_train, y_train)*100,decimals=2), np.round(lsvm_hard.score(X_test, y_test)*100, decimals=2)],
                             'acc. (PCA)': [np.round(lsvm_hard_pca.score(pca_adult, y_train)*100,decimals=2), np.round(lsvm_hard_pca.score(pca.transform(scaler_X.transform(X_test)), y_test)*100, decimals=2)]},
                            index=['training', 'test'])
display(df_lsvm_hard)

#calcolo la precisione del metodo SVM soft
C_soft = 1e1
loss = 'squared_hinge'
dual = False

lsvm_soft = LinearSVC(C=C_soft, loss=loss, dual=dual, random_state=random_seed)
lsvm_soft_pca = LinearSVC(C=C_soft, loss=loss, dual=dual, random_state=random_seed)

lsvm_soft.fit(X_train, y_train)
lsvm_soft_pca.fit(pca_adult, y_train)

df_lsvm_soft = pd.DataFrame({'acc. (no PCA)': [np.round(lsvm_soft.score(X_train, y_train)*100,decimals=2), np.round(lsvm_soft.score(X_test, y_test)*100, decimals=2)],
                             'acc. (PCA)': [np.round(lsvm_soft_pca.score(pca_adult, y_train)*100,decimals=2), np.round(lsvm_soft_pca.score(pca.transform(scaler_X.transform(X_test)), y_test)*100, decimals=2)]},
                            index=['training', 'test'])
display(df_lsvm_soft)

#testo il primo algoritmo con un kernel sigmoid, tempo medio 29 secondi 
ker_tanh = 'sigmoid'
gamma_tanh = 'auto'

C = 100

svm_tanh_1 = SVC(C=C, kernel=ker_tanh, gamma=gamma_tanh, random_state=random_seed)
svm_tanh_2 = SVC(C=C, kernel=ker_tanh, gamma=gamma_tanh, random_state=random_seed)

svm_tanh_1.fit(X_train, y_train)
svm_tanh_2.fit(pca_adult, y_train)

df_svm_tanh = pd.DataFrame({'acc. (NO PCA)': [np.round(svm_tanh_1.score(X_train, y_train)*100, decimals=2), np.round(svm_tanh_1.score(X_test, y_test)*100,decimals=2)],
                           'acc. (PCA)': [np.round(svm_tanh_2.score(pca_adult, y_train)*100,decimals=2), np.round(svm_tanh_2.score(pca.transform(scaler_X.transform(X_test)), y_test)*100,decimals=2)]},
                          index=['training', 'test'])
df_styled=df_svm_tanh.style.set_caption('SVM SIGMOID')
display(df_styled)

#testo il secondo algoritmo con un kernel rbf, tempo medio 3 minuti 38 secondi
ker_rbf = 'rbf'
gamma_rbf = 'auto'

C = 100

svm_rbf_1 = SVC(C=C, kernel=ker_rbf, gamma=gamma_rbf, random_state=random_seed)
svm_rbf_2 = SVC(C=C, kernel=ker_rbf, gamma=gamma_rbf, random_state=random_seed)

svm_rbf_1.fit(X_train, y_train)
svm_rbf_2.fit(pca_adult, y_train)

df_svm_rbf = pd.DataFrame({'acc. (NO PCA)': [np.round(svm_rbf_1.score(X_train, y_train)*100, decimals=2), np.round(svm_rbf_1.score(X_test, y_test)*100,decimals=2)],
                           'acc. (PCA)': [np.round(svm_rbf_2.score(pca_adult, y_train)*100,decimals=2), np.round(svm_rbf_2.score(pca.transform(scaler_X.transform(X_test)), y_test)*100,decimals=2)]},
                          index=['training', 'test'])
df_styled=df_svm_rbf.style.set_caption('SVM RBF')
display(df_styled)

#ora disegno i grafici per il kernel rbf, considerando solo le prime due componenti principali tempo medio 2 minuti 8 secondi
ker_rbf = 'rbf'
gamma_rbf = 'auto'

C = 100

svm = SVC(C=C, kernel=ker_rbf, gamma=gamma_rbf, random_state=random_seed)

svm.fit(pca_adult[:, :2], y_train)

df_svm_rbf = pd.DataFrame({'acc. (PCA)': [np.round(svm.score(pca_adult[:, :2], y_train)*100,decimals=2), np.round(svm.score(pca.transform(scaler_X.transform(X_test))[:, :2], y_test)*100,decimals=2)]},
                          index=['training', 'test'])

display(df_svm_rbf)

minx=min(pca_adult[:, 0].min()-1, pca_adult[:, 1].min()-1)
maxx=max(pca_adult[:, 0].max()+1, pca_adult[:, 1].max()+1)

mesh_points = 200

xx1, xx2 = np.meshgrid(np.linspace(minx, maxx, mesh_points), np.linspace(minx, maxx, mesh_points))

XX = np.concatenate([xx1.reshape(mesh_points ** 2, 1), xx2.reshape(mesh_points ** 2, 1)], axis=1)

ZZ_1 = svm.decision_function(XX)
zz_1 = ZZ_1.reshape(mesh_points, mesh_points)

plt.figure(figsize=(12, 12))
plt.scatter(pca_adult[:, 0], pca_adult[:, 1], c=y_train,marker='x', cmap=cmap, alpha=0.25)
plt.scatter(pca.transform(scaler_X.transform(X_test))[:, 0], pca.transform(scaler_X.transform(X_test))[:, 1], c=y_test,marker='o', cmap=cmap, alpha=0.25)
plt.contour(xx1, xx2, zz_1, [0], colors='red')
plt.contour(xx1,xx2,zz_1,[-1,1],colors=['green','yellow'])
plt.title('SVM RBF - DECISION BOUNDARY')

Text(0.5, 1.0, 'SVM RBF - DECISION BOUNDARY')

y_pred = svm_rbf_1.predict(X_test)
y_pred_pca = svm_rbf_2.predict(pca.transform(scaler_X.transform(X_test)))

precision = np.round(precision_score(y_test, y_pred)*100,decimals=2)
recall = np.round(recall_score(y_test, y_pred)*100,decimals=2)
fbeta = np.round(fbeta_score(y_test, y_pred, beta=1.0)*100,decimals=2)

precision_pca = np.round(precision_score(y_test, y_pred_pca)*100,decimals=2)
recall_pca = np.round(recall_score(y_test, y_pred_pca)*100,decimals=2)
fbeta_pca = np.round(fbeta_score(y_test, y_pred_pca, beta=1.0)*100,decimals=2)

df_scores = pd.DataFrame({'Precision': [precision, precision_pca], 'Recall': [recall, recall_pca], 'F1': [fbeta, fbeta_pca]}, index=['NO PCA', 'PCA'])
df_styled = df_scores.style.set_caption('SVM RBF - SCORES')
display(df_styled)

	age	workclass	fnlwgt	education	education-num	marital-status	occupation	relationship	race	sex	capital-gain	capital-loss	hours-per-week	native-country	income
0	39	State-gov	77516	Bachelors	13	Never-married	Adm-clerical	Not-in-family	White	Male	2174	0	40	United-States	<=50K
1	50	Self-emp-not-inc	83311	Bachelors	13	Married-civ-spouse	Exec-managerial	Husband	White	Male	0	0	13	United-States	<=50K
2	38	Private	215646	HS-grad	9	Divorced	Handlers-cleaners	Not-in-family	White	Male	0	0	40	United-States	<=50K
3	53	Private	234721	11th	7	Married-civ-spouse	Handlers-cleaners	Husband	Black	Male	0	0	40	United-States	<=50K
4	28	Private	338409	Bachelors	13	Married-civ-spouse	Prof-specialty	Wife	Black	Female	0	0	40	Cuba	<=50K
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
32556	27	Private	257302	Assoc-acdm	12	Married-civ-spouse	Tech-support	Wife	White	Female	0	0	38	United-States	<=50K
32557	40	Private	154374	HS-grad	9	Married-civ-spouse	Machine-op-inspct	Husband	White	Male	0	0	40	United-States	>50K
32558	58	Private	151910	HS-grad	9	Widowed	Adm-clerical	Unmarried	White	Female	0	0	40	United-States	<=50K
32559	22	Private	201490	HS-grad	9	Never-married	Adm-clerical	Own-child	White	Male	0	0	20	United-States	<=50K
32560	52	Self-emp-inc	287927	HS-grad	9	Married-civ-spouse	Exec-managerial	Wife	White	Female	15024	0	40	United-States	>50K

	age	fnlwgt	education-num	capital-gain	capital-loss	hours-per-week	workclass_ Federal-gov	workclass_ Local-gov	workclass_ Private	workclass_ Self-emp-inc	...	native-country_ Puerto-Rico	native-country_ Scotland	native-country_ South	native-country_ Taiwan	native-country_ Thailand	native-country_ Trinadad&Tobago	native-country_ United-States	native-country_ Vietnam	native-country_ Yugoslavia	income
0	39	77516	13	2174	0	40	0	0	0	0	...	0	0	0	0	0	0	1	0	0	0
1	50	83311	13	0	0	13	0	0	0	0	...	0	0	0	0	0	0	1	0	0	0
2	38	215646	9	0	0	40	0	0	1	0	...	0	0	0	0	0	0	1	0	0	0
3	53	234721	7	0	0	40	0	0	1	0	...	0	0	0	0	0	0	1	0	0	0
4	28	338409	13	0	0	40	0	0	1	0	...	0	0	0	0	0	0	0	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
32556	27	257302	12	0	0	38	0	0	1	0	...	0	0	0	0	0	0	1	0	0	0
32557	40	154374	9	0	0	40	0	0	1	0	...	0	0	0	0	0	0	1	0	0	1
32558	58	151910	9	0	0	40	0	0	1	0	...	0	0	0	0	0	0	1	0	0	0
32559	22	201490	9	0	0	20	0	0	1	0	...	0	0	0	0	0	0	1	0	0	0
32560	52	287927	9	15024	0	40	0	0	0	1	...	0	0	0	0	0	0	1	0	0	1

	age	fnlwgt	education-num	capital-gain	capital-loss	hours-per-week	workclass_ Federal-gov	workclass_ Local-gov	workclass_ Private	workclass_ Self-emp-inc	...	native-country_ Portugal	native-country_ Puerto-Rico	native-country_ Scotland	native-country_ South	native-country_ Taiwan	native-country_ Thailand	native-country_ Trinadad&Tobago	native-country_ United-States	native-country_ Vietnam	native-country_ Yugoslavia
age	1.000000	-0.076511	0.043526	0.080154	0.060165	0.101599	0.056626	0.068256	-0.210491	0.111039	...	0.001062	0.007836	0.000024	0.001923	-0.007879	-0.004940	0.007868	0.016259	-0.017775	0.000657
fnlwgt	-0.076511	1.000000	-0.044992	0.000422	-0.009750	-0.022886	-0.006932	-0.003520	0.046589	-0.025496	...	-0.014020	0.007121	-0.003055	-0.010598	0.001615	-0.001241	-0.000413	-0.083390	-0.010761	0.005707
education-num	0.043526	-0.044992	1.000000	0.124416	0.079646	0.152522	0.058244	0.097378	-0.165069	0.078843	...	-0.043058	-0.042083	0.001815	0.017812	0.049129	0.008183	-0.017134	0.127207	-0.010953	-0.001661
capital-gain	0.080154	0.000422	0.124416	1.000000	-0.032229	0.080432	-0.006299	-0.009624	-0.048185	0.096482	...	-0.003826	-0.006270	-0.002816	-0.002582	0.007639	-0.003501	-0.003603	0.012375	-0.002493	-0.002317
capital-loss	0.060165	-0.009750	0.079646	-0.032229	1.000000	0.052417	0.010380	0.014727	-0.036377	0.030956	...	-0.007343	-0.004560	-0.004175	0.005677	0.005679	-0.005191	0.008849	0.015119	0.000344	-0.005036
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
native-country_ Thailand	-0.004940	-0.001241	0.008183	-0.003501	-0.005191	0.012846	-0.004266	-0.006441	-0.008144	0.018052	...	-0.000798	-0.001430	-0.000454	-0.001154	-0.000887	1.000000	-0.000580	-0.076390	-0.001095	-0.000547
native-country_ Trinadad&Tobago	0.007868	-0.000413	-0.017134	-0.003603	0.008849	-0.007566	-0.004390	-0.001255	0.002164	0.002630	...	-0.000821	-0.001472	-0.000467	-0.001187	-0.000913	-0.000580	1.000000	-0.078606	-0.001127	-0.000563
native-country_ United-States	0.016259	-0.083390	0.127207	0.012375	0.015119	0.010673	0.017541	0.032937	-0.049809	0.007350	...	-0.108062	-0.193727	-0.061442	-0.156254	-0.120121	-0.076390	-0.078606	1.000000	-0.148334	-0.074108
native-country_ Vietnam	-0.017775	-0.010761	-0.010953	-0.002493	0.000344	-0.010381	-0.000004	-0.003952	0.009369	-0.004973	...	-0.001549	-0.002777	-0.000881	-0.002240	-0.001722	-0.001095	-0.001127	-0.148334	1.000000	-0.001062
native-country_ Yugoslavia	0.000657	0.005707	-0.001661	-0.002317	-0.005036	0.006983	-0.004139	-0.000550	0.003861	0.003343	...	-0.000774	-0.001387	-0.000440	-0.001119	-0.000860	-0.000547	-0.000563	-0.074108	-0.001062	1.000000

	Pred. Class	P(Class 0) - %	P(Class 1) - %
0	0	64.07	35.93
1	0	85.07	14.93
2	0	63.54	36.46
3	0	81.56	18.44
4	0	90.70	9.30
...	...	...	...
15076	0	69.66	30.34
15077	0	64.27	35.73
15078	1	40.30	59.70
15079	0	89.76	10.24
15080	0	99.20	0.80

	Pred. Class	P(Class 0) - %	P(Class 1) - %
0	0	51.79	48.21
1	0	78.36	21.64
2	0	68.79	31.21
3	0	78.25	21.75
4	0	83.97	16.03
...	...	...	...
15076	0	71.38	28.62
15077	1	41.91	58.09
15078	1	40.30	59.70
15079	0	88.85	11.15
15080	0	98.34	1.66

Stima del reddito annuale in base ai dati del censimento¶

Introduzione¶

Il dataset¶

1. Pulizia e visualizzazione del dataset¶

PCA¶

Introduzione

Procedimento

Varianza spiegata

MDA¶

Introduzione

Differenze tra PCA e MDA¶

PCA¶

MDA (FDA)¶

Procedimento

LDA¶

Introduzione

Procedimento

SVM Lineare¶

Introduzione

Procedimento¶

SVM Non lineare¶

Introduzione

Procedimento

	Precision	Recall	F1
NO PCA	43.360000	8.430000	14.120000
PCA	67.360000	60.910000	63.970000

	val
n. PC	79.00
expl. Var. (%)	90.51

	acc. (no PCA)	acc. (PCA)
training	79.44	83.79
test	78.93	84.09

	acc. (no PCA)	acc. (PCA)
training	79.44	83.79
test	78.93	84.09

SVM SIGMOID
	acc. (NO PCA)	acc. (PCA)
training	75.370000	77.210000
test	74.840000	77.600000